Sistem Peringatan Dini Untuk Gangguan Surge Pada Water-Cooled Chiller Berdasarkan Prediksi Nilai Vibrasi Dan Temperatur Menggunakan Metode Long Short-Term Memory (LSTM)

Fahira, Naila Najma (2024) Sistem Peringatan Dini Untuk Gangguan Surge Pada Water-Cooled Chiller Berdasarkan Prediksi Nilai Vibrasi Dan Temperatur Menggunakan Metode Long Short-Term Memory (LSTM). Diploma thesis, Institut Teknologi Sepuluh Nopember.

[thumbnail of 2040201139-Undergraduate_Thesis.pdf] Text
2040201139-Undergraduate_Thesis.pdf - Accepted Version
Restricted to Repository staff only until 1 October 2026.

Download (7MB) | Request a copy

Abstract

Dalam industri farmasi, menjaga kualitas produk dengan mengatur kondisi lingkungan seperti suhu, kelembapan, dan tekanan sangatlah penting. Chiller, alat pendingin utama dalam operasi manufaktur, beroprasi sejak 1996 telah mengalami penurunan performa seiring berjalannya waktu. Salah satu masalah yang sering terjadi adalah surge, dimana ada perbedaan tekanan pada kompresor melebihi set point, menyebabkan aliran refrigeran berbalik dari kondensor ke kompresor, yang dapat merusak chiller. Saat ini, pemeliharaan chiller dilakukan secara preventif dengan pemeriksaan secara periodik. Namun, pendekatan ini dianggap kurang optimal karena tidak mampu mengantisipasi gangguan yang memiliki waktu tidak terduga. Untuk mengatasi masalah ini, dikembangkan sebuah sistem pemeliharaan prediktif menggunakan sensor vibrasi 3 axis dan temperatur pada motor kompresor chiller sebagai komponen utama yang dipantau secara real-time. Sistem ini memanfaatkan time-series forecasting untuk memprediksi nilai sensor pada waktu mendatang dengan tujuan dapat melakukan tindakan pemeliharaan sebelum mencapai nilai threshold, menggunakan metode Long Short-Term Memory (LSTM). Dataset diproses melalui tahap exponensial smoothing dan normalisasi data. Hasil dari penelitian menunjukkan bahwa variabel yang digunakan dalam early warning chiller adalah temperatur, kecepatan getaran sumbu-z, dan percepatan getaran sumbu-x. Model terbaik dalam memprediksi nilai keluaran sensor yaitu menggunakan satu hidden layer yang memiliki 96 neuron, input dan output prediksi sebanyak 96 data per variabel, dan dilatih selama 150 epoch. Model tersebut mencapai rata-rata akurasi pada MAPE sebesar 1,45% dan RMSE sebesar 0,04 yang termasuk kategori nilai akurasi sangat baik. Dengan sistem ini, dapat diprediksi nilai keluaran sensor untuk 48 jam ke depan, yang dapat digunakan sebagai early warning system dan memudahkan teknisi untuk merencanakan perawatan sebelum gangguan surge terjadi.
==================================================================================================================================
In the pharmaceutical industry, maintaining product quality by controlling environmental conditions such as temperature, humidity, and pressure is crucial. Chillers, the main cooling equipment in manufacturing operations, have been in use since 1996 and have experienced performance degradation over time. One common issue is surge disturbance, where there is a pressure difference in the compressor exceeding the set point, causing refrigerant flow to reverse from the condenser to the compressor, potentially damaging the chiller. Currently, chiller maintenance is performed preventively through periodic inspections. However, this approach is considered suboptimal as it cannot anticipate surge disturbances with unpredictable timing. To address this issue, a predictive maintenance system has been developed using 3-axis vibration and temperature sensors on the chiller compressor motor as key real-time monitoring components. This system utilizes time-series forecasting to predict sensor values in the future, aiming to take maintenance actions before reaching threshold values, employing Long Short-Term Memory (LSTM) methods. The dataset undergoes exponential smoothing and data normalization stages. Results from the study indicate that variables used in the chiller's early warning system include temperature, z-axis vibration velocity, and x-axis vibration acceleration. The best model for predicting sensor output values uses a single hidden layer with 96 neurons, predicting 96 data points per variable in and out, trained over 150 epochs. This model achieved an average Mean Absolute Percentage Error (MAPE) accuracy of 1.45% and Root Mean Square Error (RMSE) of 0.04, classifying it as having excellent accuracy. With this system, sensor output values can be predicted up to 48 hours in advance, serving as an early warning system and facilitating technicians in planning maintenance before surge disturbances occur.

Item Type: Thesis (Diploma)
Uncontrolled Keywords: Pemeliharaan Prediktif, Surge, Temperatur, Time-series Forecasting, Vibrasi, Chiller, Early Warning System, Long Short-Term Memory, Predictive Maintenance, Surge, Temperature, Vibration
Subjects: T Technology > T Technology (General) > T174 Technological forecasting
T Technology > T Technology (General) > T385 Visualization--Technique
T Technology > T Technology (General) > T57.5 Data Processing
T Technology > TJ Mechanical engineering and machinery > TJ174 Maintenance and repair of machinery
T Technology > TJ Mechanical engineering and machinery > TJ213 Automatic control.
T Technology > TJ Mechanical engineering and machinery > TJ217.6 Predictive Control
T Technology > TJ Mechanical engineering and machinery > TJ990 Compressors
T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK3070 Automatic control
T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK351 Electric measurements.
T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK4055 Electric motor
T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5103.2 Wireless communication systems. Two way wireless communication
T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5105.888 Web sites--Design. Web site development.
T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7878 Electronic instruments
Divisions: Faculty of Vocational > 36304-Automation Electronic Engineering
Depositing User: Naila Najma Fahira
Date Deposited: 21 Aug 2024 04:49
Last Modified: 21 Aug 2024 04:49
URI: http://repository.its.ac.id/id/eprint/115487

Actions (login required)

View Item View Item